CLean up a bit

create_handler.ipynb

@@ -21,8 +21,9 @@
     }
    ],
    "source": [
-    "%%writefile requirements.txt\n",
-    "git+https://github.com/openai/whisper.git@8cf36f3508c9acd341a45eb2364239a3d81458b9"
+    "% % writefile requirements.txt\n",
+    "git + https: // github.com / openai / whisper.git @ 8\n",
+    "cf36f3508c9acd341a45eb2364239a3d81458b9"
    ]
   },
   {
@@ -65,7 +66,7 @@
     }
    ],
    "source": [
-    "!wget https://cdn-media.huggingface.co/speech_samples/sample1.flac"
+    "!wget https: // cdn-media.huggingface.co / speech_samples / sample1.flac"
    ]
   },
   {
@@ -118,8 +119,8 @@
     }
    ],
    "source": [
-    "%%writefile handler.py\n",
-    "from typing import  Dict\n",
+    "% % writefile handler.py\n",
+    "from typing import Dict\n",
     "from transformers.pipelines.audio_utils import ffmpeg_read\n",
     "import whisper\n",
     "import torch\n",
@@ -127,12 +128,10 @@
     "SAMPLE_RATE = 16000\n",
     "\n",
     "\n",
-    "\n",
     "class EndpointHandler():\n",
     "    def __init__(self, path=\"\"):\n",
     "        # load the model\n",
-    "        self.model = whisper.load_model(\"medium\")\n",
-    "\n",
+    "        self.model = whisper.load_model(\"small\")\n",
     "\n",
     "    def __call__(self, data: Dict[str, bytes]) -> Dict[str, str]:\n",
     "        \"\"\"\n",
@@ -145,17 +144,25 @@
     "        # process input\n",
     "        inputs = data.pop(\"inputs\", data)\n",
     "        audio_nparray = ffmpeg_read(inputs, SAMPLE_RATE)\n",
-    "        audio_tensor= torch.from_numpy(audio_nparray)\n",
-    "        \n",
+    "\n",
     "        # run inference pipeline\n",
     "        result = self.model.transcribe(audio_nparray)\n",
     "\n",
     "        print(\"Hi this is a custom log!\")\n",
     "\n",
     "        # postprocess the prediction\n",
-    "        return { \"text\": result[\"text\"] }"
+    "        return {\"text\": result[\"text\"]}"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "outputs": [],
+   "source": [],
+   "metadata": {
+    "collapsed": false
+   }
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -204,8 +211,7 @@
     "\n",
     "# file reader\n",
     "with open(\"sample1.flac\", \"rb\") as f:\n",
-    "  request = {\"inputs\": f.read()}\n",
-    "\n",
+    "    request = {\"inputs\": f.read()}\n",
     "\n",
     "# test the handler\n",
     "pred = my_handler(request)"
@@ -250,7 +256,8 @@
    "source": [
     "import json\n",
     "\n",
-    "json.dumps({'transcription': \" going along slushy country roads and speaking to damp audiences in draughty school rooms day after day for a fortnight. He'll have to put in an appearance at some place of worship on Sunday morning, and he can come to us immediately afterwards.\"})"
+    "json.dumps({\n",
+    "    'transcription': \" going along slushy country roads and speaking to damp audiences in draughty school rooms day after day for a fortnight. He'll have to put in an appearance at some place of worship on Sunday morning, and he can come to us immediately afterwards.\"})"
    ]
   },
   {

handler.py

@@ -1,4 +1,4 @@
-from typing import  Dict
+from typing import Dict
 from transformers.pipelines.audio_utils import ffmpeg_read
 import whisper
 import torch
@@ -6,12 +6,9 @@ import torch
 SAMPLE_RATE = 16000
 
 
-
-class EndpointHandler():
+class EndpointHandler:
     def __init__(self, path=""):
-        # load the model
-        self.model = whisper.load_model("medium")
-
+        self.model = whisper.load_model("small")
 
     def __call__(self, data: Dict[str, bytes]) -> Dict[str, str]:
         """
@@ -24,12 +21,11 @@ class EndpointHandler():
         # process input
         inputs = data.pop("inputs", data)
         audio_nparray = ffmpeg_read(inputs, SAMPLE_RATE)
-        audio_tensor= torch.from_numpy(audio_nparray)
-        
+
         # run inference pipeline
         result = self.model.transcribe(audio_nparray)
-        
+
         print("Hi this is a custom log!")
 
         # postprocess the prediction
-        return { "text": result["text"] }
+        return {"text": result["text"]}